Neuroplasticity of excitatory and inhibitory cortical circuits in schizophrenia.

Schizophrenia is a neurodevelopmental disorder characterized by deficits in cognitive processes mediated by the circuitry of the dorsolateral prefrontal cortex (DLPFC). These deficits are associated with a range of alterations in DLPFC circuitry, some of which reflect the pathology of the illness and others of which reflect the neuroplasticity of the brain in response to the underlying disease process. This article reviews disturbances in excitatory and inhibitory components of DLPFC circuitry from the perspective of developmental neuroplasticity and discusses their implications for the identification of novel therapeutic targets.

Neuroplasticity can be broadly considered to be the capacity of the brain to change the molecular and structural features that dictate its functions in response to a disease process (or other factors) that disrupts those functions.[1] For a disorder such as schizophrenia, the disease process appears to result from a complex interplay of an unknown number of genetic liabilities and environmental risk factors that unleash pathogenetic mechanisms which produce a pathological entity, a conserved set of molecular and cellular disturbances in specific neural circuits. These pathological changes so alter the normal function of the affected circuits that the resulting pathophysiology gives rise to the emergent properties recognized as the clinical features of the illness.[2] One approach to dissecting this disease process involves focusing on a well-defined clinical component of the illness. For example, deficits in cognitive abilities are thought to be the core features of schizophrenia because they occur with high frequency in individuals with schizophrenia, are relatively stable over the course of the illness, are independent of the psychotic symptoms of the disorder, are present in a milder form in individuals at genetic risk who do not become clinically ill,[3] and are the best predictor of long-term functional outcome.[4]

Of the domains of cognition affected in schizophrenia, disturbances in working memory, the ability to transiently maintain and manipulate a limited amount of information in order to guide thought or behavior, are accompanied by altered activation of the dorsolateral prefrontal cortex (DLPFC, The altered activation of the DLPFC under such conditions might be specific to the disease process of schizophrenia because these disturbances are present in medication-naïve individuals with schizophrenia, but not in subjects with other psychotic disorders or major depression.[5,6]

This review examines alterations in components of excitatory and inhibitory neurotransmission in DLPFC circuitry that might contribute to the impairments in working memory in schizophrenia. Each mediator is considered from the perspective of which alterations reflect the disease process and which might be neuroplastic responses of the affected circuits. Although additional studies are required, existing data suggests that many of the alterations described below are probably also present in other cortical regions that are dysfunctional in schizophrenia.[7]

Neuroplasticity of excitatory cortical connections in schizophrenia

Excitatory connections in the DLPFC are altered in schizophrenia

The disease process of schizophrenia appears to involve deficient glutamate -mediated excitatory neurotransmission through the N-methyl-D-aspartic acid (NMDA) receptor.[8,9] NMDA receptor antagonists such as phencyclidine (PCP) or ketamine increase both positive and negative symptoms in patients with schizophrenia, and the administration of subanesthetic doses of ketamine to healthy individuals produces thought disorder and other features similar to those seen in schizophrenia.[10]

In addition, systemic administration of NMDA receptor antagonists disrupts working memory in rats,[11] and application of an NMDA receptor antagonist to the DLPFC impairs working memory performance in monkeys.[12] However, although postmortem studies have reported alterations in measures of glutamate receptor binding, transcription, and subunit protein expression in several brain regions in subjects with schizophrenia,[13] such findings for mRNA and protein levels of NMDA receptor subunits in the DLPFC have been limited in magnitude and not always replicated, suggesting that other components of NMDA receptor signaling might be affected in the illness.[14]

Anatomical studies do support the presence of input-specific alterations of excitatory connections in the DLPFC in schizophrenia. In the DLPFC, pyramidal neurons ( are the principal source of glutamate neurotransmission, as well as the targets of the majority of glutamate-containing axon terminals. Although the number of these neurons does not appear to be altered in schizophrenia,[15,16] neuronal density in the DLPFC has been reported to be increased in schizophrenia.[17] Increased cell packing density has been interpreted as evidence of a reduction in the amount of cortical neuropil, the axon terminals, dendritic spines, and glial processes that occupy the space between neurons.[18] Consistent with this interpretation, synaptophysin protein, a marker of axon terminals, has been reported to be decreased in the DLPFC of subjects with schizophrenia.[19,21] Furthermore, gene expression profiling studies have found reduced tissue levels of gene transcripts that encode proteins involved in the presynaptic regulation of neurotransmission.[22]

Dendritic spines are the principal targets of excitatory synapses to pyramidal neurons. Although most dendritic spines present are stable in number during adulthood,[23] they are subject to a number of neuroplastic changes, such as a loss of their presynaptic excitatory input. In schizophrenia, dendritic spine density in pyramidal neurons has been reported to be lower in the DLPFC[24,25]; understanding the nature of these neuroplastic responses requires knowledge of the specific circuits that are affected and the developmental mechanisms that might underlie these changes.

Reduced excitatory connections in schizophrenia are specific to a subset of pyramidal neurons

Pyramidal neurons can be divided into subgroups based on the brain region targeted by their principal axonal projection and the sources of their excitatory inputs; both of these characteristics are associated with the location of pyramidal cell bodies in different layers of the cortex ( For example, many pyramidal cells in layers 2 to 3 send axonal projections to other cortical regions, pyramidal neurons in layer 5 tend to project to the striatum and other subcortical structures, and pyramidal neurons in layer 6 furnish projections primarily to the thalamus.[26] Studies of basilar dendritic spine density on Golgi-impregnaled pyramidal neurons in each cortical layer of the DLPFC in the same cohort of subjects found a significant effect of diagnosis on spine density only for pyramidal neurons in deep layer 3

The functional integrity of the pyramidal neurons with lower dendritic spine densities may be reflected in changes in their somal volume. For example, shifts in somal size may indicate disturbances in neuronal connectivity, given that somal size has been shown to be correlated with measures of a neuron's dendritic tree[28] and axonal arbor.[29] Indeed, the mean cross-sectional somal area of the Golgi-impregnated, deep layer 3 pyramidal neurons was 9% smaller in the subjects with schizophrenia relative to normal control subjects.[25] Consistent with this observation, the mean somal volume of Nisslstained pyramidal neurons in DLPFC deep layer 3 was also 9% smaller in a different cohort of subjects with schizophrenia.[30] Similarly, in another study, the mean somal size of all layer 3 neurons in DLPFC area 9 was smaller in subjects with schizophrenia, and was accompanied by a decrease in the density of the largest neurons in deep layer 3, without a change in somal volume in layer 5.[31] Furthermore, in both primary and association auditory cortices, somal volumes of deep layer 3, but not of layer 5, pyramidal neurons were smaller in schizophrenia.[32,33] Together, these findings suggest that in schizophrenia: i) basilar dendritic spine density is lower and somal volume is smaller in deep layer 3 pyramidal neurons; ii) these alterations are specific to or at least most prominent in deep layer 3; iii) this pattern of alterations is not restricted to the DLPFC; and iv) these differences reflect the underlying disease process and not confounding factors.

The contribution of developmental plasticity to dendritic spine alterations in schizophrenia

Dendritic spine density on DLPFC layer 3 pyramidal neurons undergoes a substantial decline during adolescence in primates.[34] Consistent with the findings that dendritic spines are the main site of excitatory synaptic input onto pyramidal cells and that all mature dendritic spines contain an excitatory synapse,[35] the number of excitatory synapses declines in a similar age-related fashion in both monkey and human DLPFC.[36,37] In humans, this synaptic pruning is thought to underlie the decrease in cortical gray matter thickness that occurs during adolescence.[38,39] Interestingly, the late developmental refinements in excitatory connectivity are more marked in layer 3 than in the deeper cortical layers,[36] suggesting that they may be associated with the apparent laminas-specific alterations in spine density in schizophrenia. The observation of alterations in the expression of certain synaptic proteins in schizophrenia suggested the possibility that the exuberant synapses present before adolescence somehow compensated for a dysfunction in excitatory transmission in individuals with schizophrenia.[40] Alternatively, such alterations in synaptic protein expression might disturb the mechanisms of adolescence-related synapse elimination leading, for instance, to excessive synapse pruning and decreased spine number in the illness.[41,42]

The potential contribution of excitatory synapse pruning during adolescence to disease-related changes in DLPFC function depends, in part, on the functional properties of the synapses that are pruned. During early brain development, pruned synapses are functionally immature. Immature glutamate synapses are relatively weak and their maturation involves an activity-dependent increase in strength. Such activity-dependent strengthening might underlie synapse stabilization, and thus mark for elimination the immature synapses that are not strengthened.[43,44] However, recent findings in the developing monkey DLPFC indicate that the excitatory-inputs to layer 3 pyramidal neurons mature functionally during the age range when they are present in high density and before synaptic pruning begins.[45] Thus, these data suggest that the substantial remodeling of excitatory connectivity of the primate DLPFC during adolescence primarily involves the elimination of mature synapses, and that some other factor, such as the neuronal source of input, somehow tags mature synapses for pruning.[46] Thus, the presence of functionally mature synapses prior to adolescence supports the hypothesis that the excess in excitatory synapse number prior to adolescence might be able to compensate for a molecular based dysfunction of these synapses in individuals with schizophrenia, and thereby forestall the appearance of the clinical features of the illness until synapse number falls below some critical threshold.[40]

Neuroplasticity of inhibitory cortical connections in schizophrenia

Prefrontal inhibitory neurotransmission is altered in schizophrenia

Studies from multiple laboratories have consistently found lower levels of the mRNA for the 67 kilodalton isoform of glutamic acid decarboxylase (GAD67), the principal synthesizing enzyme for y-aminobutyric acid (GABA), in the DLPFC of subjects with schizophrenia.[16,22-47,52] At the cellular level, the expression of GAD67 mRNA was not detectable in -25% to 35% of GABA neurons in layers 15 of the DLPFC, but the remaining GABA neurons exhibited normal levels of GAD67 mRNA.[16-47] Similarly, expression of the mRNA for the GABA membrane transporter (GAT1), a protein responsible for reuptake of released GABA into nerve terminals, was decreased in a similar minority of GABA neurons.[53] These findings suggest that both the synthesis and reuptake of GABA are lower in a subset of DLPFC neurons in schizophrenia.

Subclasses of cortical GABA neurons can be distinguished on the basis of a number of molecular, electro-physiological, and anatomical properties. For example, the affected GABA neurons in schizophrenia include the subclass that contain the calcium-binding protein, parvalbumin (PV), which comprise -25% of GABA neurons in the primate DLPFC. PV-containing neurons include fast-spiking chandelier and basket neurons that principally target the axon initial segments and cell body/proximal dendrites, respectively, of pyramidal neurons[54,55] ( In individuals with schizophrenia the expression level of PV mRNA is reduced, although the number of PV neurons appears to be unchanged56; in addition, approximately half of PV mRNA-containing neurons lack detectable levels of GAD67 mRNA.[57] In contrast, the -50% of GABA neurons that express the calcium binding protein calretinin appear to be unaffected.[57]

In the DLPFC of subjects with schizophrenia, G ATI immunore activity is selectively reduced in the characteristic axon terminals (cartridges) of PV-containing chandelier neurons.[58] In the postsynaptic targets of these axon cartridges, the axon initial segments of pyramidal neurons, immunoreactivity for the GABAA receptor α2 subunit (which is present in most GABAA receptors in this location[59]) is markedly increased in schizophrenia.[60] Several lines of evidence suggest that the reductions in presynaptic GABA markers (GAT1 and PV) and increased postsynaptic GABAA receptors are compensatory responses to a deficit in GABA release from chandelier neurons. For example, PV is a slow calcium buffer that does not affect the amplitude, but accelerates the decay, of Ca2+ transients in GABA nerve terminals.[61,62] Thus, PV decreases the residual Ca2+ levels that normally accumulate in nerve terminals and facilitate GABA release during repetitive firing.[61] Studies in PV-deficient mice have demonstrated that a decrease in PV increases residual Ca2+ and favors synaptic facilitation.[61,63] Furthermore, the enhanced facilitation of GABA release from fast-spiking neurons with reductions in PV is associated with increased power of gamma oscillations[63] (which is, as explained below, deficient in schizophrenia). Similarly, the blockade of GABA reuptake via GAT1 prolongs the duration of inhibitory postsynaptic currents (IPSCs) when synapses located close to each other are activated synchronously[64]; the resulting prolongation of IPSCs increases the probability of IPSC summation, enhances the total efficacy of IPSC trains, and thereby augments GABA signaling. The upregulation of the postsynaptic GABAA receptors that contain α2, subunits would be expected to increase the efficacy of the GABA that is released from chandelier neurons. Thus, the combined reduction of PV and GAT1 proteins in chandelier cell axon cartridges, and the upregulation of postsynaptic GABAA receptors, appear to represent neuroplastic responses that might act synergistically to increase the efficacy of GABA neurotransmission at pyramidal neuron axon initial segments during the types of repetitive neuronal activity associated with working memory.

However, the persistence of cognitive impairments in individuals with schizophrenia suggests that these neuroplastic changes in GABA neurotransmission from chandelier neurons are insufficient as compensatory responses. Alternatively, it is possible that compensation at chandelier cell synapses is not effective because additional interneuron subclasses are also functionally deficient in schizophrenia.[65] Consistent with this interpretation, other findings indicate that alterations in PVcontaining GABA neurons cannot account for all of the observed findings in postmortem studies of schizophrenia. For example, the levels of GAD67 and GAT1 mRNAs are reduced to comparable degrees in layers 2-5 ,[47,53] even though the density of PV neurons is much greater in layers 3 and 4 than in layers 2 and 5.[66] In addition, PV mRNA expression was reduced in layers 3 and 4, but not in layers 2 and 5, in subjects with schizophrenia.[57] Indeed, other studies have found lower tissue concentrations of the mRNAs for the neuropeptides somatostatin (SST) and cholecystokinin (CCK) in the DLPFC of subjects with schizophrenia ( [51]

In the cortex, SST is expressed by GABA neurons located in layers 2 and 5 that do not express PV or CR.[67] CCK is also heavily expressed in GABA neurons that do not contain either PV or SST located principally in layers 2-3 of the primate prefrontal cortex.[68]

Interestingly, the axon terminals of CCK-containing large basket neurons, which target selectively pyramidal neuron cell bodies, contain type I cannabinoid receptors (CB1R),[69] and the mRNA and protein levels of CB1R are also lower in schizophrenia.[70] Because activation of the CB1R suppresses GABA release from the terminals of CCK neurons, the downregulation of this receptor may represent a compensatory response to reduce the ability of endogenous cannabinoids to decrease GABA release from CCK/CBlR-containing axon terminals.[70]

Altered GABA neurotransmission in PV-containing neurons impairs prefrontal network synchrony in schizophrenia

Reduced GABA signaling from PV-containing GABA neurons to the perisomatic region of pyramidal neurons in the DLPFC might contribute to the pathophysiology of working memory dysfunction via the following mechanisms. First, the activity of DLPFC GABA neurons is essential for normal working memory function in monkeys.[71,72] Second, PV-positive GABA neurons and pyramidal neurons share common sources (eg, thalamic afferents) of excitatory input.[55] The resulting feed-forward, disynaptic inhibition creates a time window during which the number of excitatory inputs required to evoke pyramidal neuron firing must occur.[73] Third, both chandelier and basket neurons target multiple pyramidal neurons,[74] enabling them to use this timing mechanism to synchronize the activity of local populations of pyramidal neurons.[75] Fourth, networks of PV-positive GABA neurons, formed by both chemical and electrical synapses, give rise to oscillatory activity in the gamma band range, the synchronized firing of a neuronal population at 30 to 80 Hz.[76,77] Interestingly, gamma band oscillations in the human DLPFC increase in proportion to working memory load,[78] and in subjects with schizophrenia, prefrontal gamma band oscillations are reduced bilaterally during a working memory task.[79] Thus, a deficit in the synchronization of pyramidal cell firing, resulting from impaired regulation of pyramidal cell networks by PV-positive GABA neurons, may contribute to reduced levels of induced gamma band oscillations, and consequently to impairments in cognitive tasks that involve working memory in subjects with schizophrenia.[65] Interestingly, CCK/CB1R- and PV-containing cells provide convergent sources of perisomatic inhibition to pyramidal neurons that play specific roles in shaping network activity, including complementary roles in regulating gamma band oscillations.[80] Thus, alterations in CCK-containing basket cells could also contribute to impaired gamma oscillations in schizophrenia.

The contribution of developmental plasticity to GABA neuron alterations in schizophrenia

In the monkey prefrontal cortex DLPFC, the density of symmetric, presumably GABA, synapses rises rapidly during the third trimester of gestation and perinatal period until stable, adult levels are achieved at 3 months postnatal.[36] In contrast, pre- and postsynaptic markers of the functional properties of chandelier axon inputs to the axon initial segment (AIS) of pyramidal neurons exhibit a very protracted maturation. Presynaptically, immunoreactivities for the calcium-binding protein PV and GAT1 in chandelier axon cartridges are not detectable or low at birth, rise (albeit with different developmental time courses) to peak levels early in postnatal development that are sustained until -15 months of age, and then rapidly decline during adolescence until stable adult levels are achieved.[34,81,82] Since chandelier cartridges are readily visualized with Golgi staining across postnatal development,[83] these changes in PV and GAT1 immunoreactivity are likely to reflect shifts in the concentration of these proteins rather than changes in the presence of, or in the density of, axon terminals within chandelier axon cartridges.[82] Post-synaptically, GABAA receptors containing α2 subunits predominate in pyramidal neuron AIS especially in cortical layers 2-4.[84] The density of pyramidal neuron AIS immunoreactive for oc2 subunits is high at birth, then significantly declines during adolescence before achieving stable adult levels:[82] These findings indicate that both pre- and postsynaptic markers of GABA neurotransmission undergo significant changes during postnatal development, suggesting that the capacity to synchronize pyramidal neuron output in the prefrontal cortex (PFC) might be in substantial flux until adulthood.

The significance of these changes depends, at least in part, on how GABA synapses are stabilized at the AIS while the functional properties of GABA neurotransmission at this location are changing during postnatal development. For example, pyramidal neuron AIS contain specific proteins that regulate synapse structure and receptor clustering: i) Ankyrin-G, an adaptor molecule that links various membrane proteins to the cytoskeleton, is localized to AIS,[85] and interactions between ankyrin-G and the cell adhesion molecule, neurofascin, are required for the formation and stabilization of GABA synapses at AIS86; ii) piV spectrin, which is localized to the AIS of pyramidal neurons through its direct interaction with ankryin-G,87 is a critical component in the organization and stabilization of membrane proteins at the AIS88; iii) The scaffolding protein gephyrin facilitates the preferential accumulation of gephyrin-GABAA receptor clusters, especially those containing α2 subunits.[89] In monkey DLPFC, ankyrin-G and βiV spectrin-labeled AIS decline in density during the first 6 months postnatal, but then remain stable, whereas the density of gephyrin-labeled AIS is stable through early postnatal development and then then markedly declines during adolescence.[90] Thus, molecular determinants of the structural features that define GABA inputs to pyramidal neuron AIS in monkey PFC undergo distinct developmental trajectories, with different types of changes occurring during the perinatal period and adolescence (

This complex and protracted postnatal maturation of the inputs from PV-containing GABA neurons in the primate DLPFC provides a number of opportunities for even subtle disturbances to have their effects amplified as they alter the trajectories of the developmental events that follow. For example, the marked developmental changes in the axon terminals of PV-containing chandelier neurons, and their postsynaptic receptors, during the perinatal period and adolescence, raises the possibility that the alterations in schizophrenia of these markers reflect a disturbance in these patterns of development.

These temporal correlations may explain how a range of environmental factors (eg, labor-delivery complications, urban place of rearing, and marijuana use during adolescence) are all associated with increased risk for the appearance of schizophrenia later in life. Although speculative, the current literature raises the possibility that the GABA-related disturbances in schizophrenia represent an arrest of normal developmental trajectories. For example, recent studies indicate that mRNA expression for the al subunit of the GABAA receptor in the DLPFC increases across postnatal development, whereas the expression of the α2 subunit declines.[91] Thus, the elevated α2 60 subunit expression[60] and the decreased α1 mRNA levels92 in schizophrenia might reflect a developmental dysregulation of GABAA receptor a subunit expression, where the changes in subunit expression with age fail to undergo their full course. Under this scenario, the alterations of DLPFC circuitry in schizophrenia may render it unable to support higher levels of working memory load, rendering the impaired performance in schizophrenia analogous to the immature levels of working memory function seen in children.[93,94]

Neuroplastic responses as targets for treatment

The findings reviewed above indicate that working memory and related cognitive impairments in schizophrenia are likely the result of a complex set of alterations in prefrontal excitatory and inhibitory circuitry. Some of these alterations appear to be deleterious causes or consequences of disturbances in the functional architecture of the DLPFC and interconnected brain regions, whereas others may be best explained as compensatory responses. In each case, they reflect the morphological and molecular neuroplasticity of DLPFC circuitry in a disease state. Understanding whether the disease-related change in a given molecule is a consequence or compensation in the disease process has important implications both for the nature of activity of the drugs designed against that target and for the potential therapeutic value of the target. For example, is the neuroplastic capacity of cortical circuitry sufficiently limited that pharmacological augmentation of a compensatory response is feasible? The results of a recent proofof -concept clinical trial suggest that this may be the case. For example, the idea that GABAA receptors α2 subunits are upregulated in pyramidal neurons due to a deficit in GABA input from chandelier neurons led to the use of a novel, positive allosteric modulator of this receptor subtype that improved both working memory function and prefrontal gamma band oscillations in a small randomized controlled trial of subjects with schizophrenia.[95] Given the marked developmental changes that occur in each of these systems during adolescence, this type of pharmacological intervention may have particular value as a treatment strategy for high-risk adolescents in the prodromal phase of the illness. However, the effectiveness and safety of such interventions requires a fuller understanding of the maturation of these neural circuits, of the functional consequences of these circuitry changes and of the vulnerability of these developmental processes to pharmacological agents.